4 ms·

This is a natural language library, and languages still care very much about gender.

by revelation 10y ago

dack 10y ago
but can a function which is basically "name -> gender" really be honest with its users? It's claiming something that would require much more information to predict with any reasonable chance of success (even excluding lgbt folks). If you look at the files it is based on, it looks like there are only a tiny number of names defined anyway - what hope does the user have of accomplishing the goal they want? Now, I might accept an API in which machine learning was involved and tons of historical data was included as input before predicting the value. Even then, it would of course have a decent potential of being wrong, but at least the result would be a more realistic claim. The current API feels a lot like this sort of thing: boolean isHappySentence(String sentence) { return sentence.contains(":)"); } Sure, it might get it right some of the time, but it's not effective enough to be useful.
- komali2 10y ago
  Every single aspect of this API has a degree of uncertainty. If you'd like to improve the capability of the algorithm, put in some PRs to help out the accuracy.
spraak 10y ago
Right, but I think that they're pointing out that it's important to understand that gender is much more than a grammatical construct. It's not even just he/she as some people identify e.g. as fluid or other non binary genders
dustinmr 10y ago
If you get outside of English and think of romance languages, every noun has a gender, and it changes the words you use based on that gender. Its a typical mistake non-native, and even native speakers, make in Spanish. Definitely one of the things I still regularly get wrong after 9 years of speaking the language daily. Which makes it potentially very important to trying to machine process a language.

4 ms·

This is a natural language library, and languages still care very much about gender.

by revelation 10y ago

This is a natural language library, and languages still care very much about gender.

dack 10y ago
but can a function which is basically "name -> gender" really be honest with its users? It's claiming something that would require much more information to predict with any reasonable chance of success (even excluding lgbt folks). If you look at the files it is based on, it looks like there are only a tiny number of names defined anyway - what hope does the user have of accomplishing the goal they want? Now, I might accept an API in which machine learning was involved and tons of historical data was included as input before predicting the value. Even then, it would of course have a decent potential of being wrong, but at least the result would be a more realistic claim. The current API feels a lot like this sort of thing: boolean isHappySentence(String sentence) { return sentence.contains(":)"); } Sure, it might get it right some of the time, but it's not effective enough to be useful.
- komali2 10y ago
  Every single aspect of this API has a degree of uncertainty. If you'd like to improve the capability of the algorithm, put in some PRs to help out the accuracy.
spraak 10y ago
Right, but I think that they're pointing out that it's important to understand that gender is much more than a grammatical construct. It's not even just he/she as some people identify e.g. as fluid or other non binary genders
dustinmr 10y ago
If you get outside of English and think of romance languages, every noun has a gender, and it changes the words you use based on that gender. Its a typical mistake non-native, and even native speakers, make in Spanish. Definitely one of the things I still regularly get wrong after 9 years of speaking the language daily. Which makes it potentially very important to trying to machine process a language.